Download Chapter 5 Niche differences between sexual and apomictic

Chapter 5
Chapter 5
Niche differences between sexual and apomictic
Taraxacum as a consequence of both ploidy
effects and selection
59
Niche differences between sexual and apomictic
Taraxacum as a consequence of both ploidy effects and
selection
Carolien G.F. de Kovel
Abstract
Multicellular organisms are usually sexual, but in some species asexual
genotypes can spin off from the sexual population. In most cases, these asexual
genotypes are polyploid. By comparing such newly spun off asexual genotypes with
established asexual genotypes and with the sexual ancestors, the effects of selection
and polyploidy on differences between sexual and asexual conspecifics can be
disentangled. In Taraxacum officinale the triploid apomicts had larger cells than the
diploid sexuals because of the ploidy difference. Polyploidy had a negative effect on
leaf number. Apomicts were selected for higher proportion of viable seeds. Despite
positive heritabilities for seed weight and number of ovules per capitulum, no
directional selection on these traits was noticeable. Selection increased plasticity in
leaf length response to shading in the apomicts. The consequences of ploidy and
selection effects for the stability of the mixed sexual-asexual system are discussed.
Keywords: apomixis, niche, polyploidy, selection, sexual reproduction,
Taraxacum
60
Chapter 5
Introduction
Multicellular organisms usually reproduce by way of sex. Yet, in a number of
species asexual forms coexist with the sexual forms. These asexual forms have arisen
secondarily from sexual forms, quite often through hybridisation (Bierzychudek
1985;Butlin et al. 1999). As a result of this hybridisation, the asexual forms are
polyploid in many instances (Bierzychudek 1985;Suomalainen et al. 1987). Studying
the significance of sex by comparing sexual and asexual forms of the same species is
often complicated by this difference in ploidy level.
In some species, new asexual genotypes arise regularly. In the case of
Taraxacum, dandelion, new asexual genotypes can arise from crosses between
existing apomictic plants and sexual plants, which occur in mixed sexual-apomictic
populations. In Taraxacum, the apomictic genotypes are triploid, whereas the sexual
genotypes are diploid. The origin of triploid apomictic dandelions is unknown, though
it seems clear that hybridisation between related species or races played a role
(Richards 1973). Contemporary apomictic genotypes of Taraxacum are probably
formed from backcrosses between existing apomictic genotypes and sexual genotypes
(Menken et al. 1995;Morita et al. 1990;Morita et al. 1990). Though apomictic
Taraxacum produces seed parthenogenetically, it produces pollen through meiotic
division. Most of this pollen is sterile, but some grains are able to fertilise a haploid
egg-cell of a sexual plant. If this pollen is diploid, a new triploid will be formed that is
usually apomictic (Tas & Van Dijk 1999a;Tas & Van Dijk 1999b;Den Nijs & Menken
1994). Sexuals and apomicts occur in mixed populations (Den Nijs & Sterk
1984b;Den Nijs & Sterk 1984a), and from allozyme data it is likely that new apomicts
arise from the local population (Menken et al. 1995). This system allows us to study
the dynamics of sexual and asexual forms more closely.
Because of the differences in ploidy level between sexuals and apomicts,
direct comparison may reveal differences that are either a direct consequence of
polyploidisation, or that are the result of different selection regimes on sexuals and
apomicts. By studying new apomictic genotypes that have not encountered much
selection together with apomictic and sexual genotypes, we may be able to disentangle
the effects of polyploidy and selection.
New triploid genotypes of Taraxacum were generated by placing sexual plants
in a field containing only apomicts, so all offspring was fathered by apomicts. Seeds
were taken to the lab and screened on ploidy level with a flow-cytometer (Ulrich &
Ulrich 1991). Triploid offspring plants were selected to provide seeds for this
experiment. These new or hybrid genotypes have encountered little selection. If these
genotypes closely resemble the established apomicts in some traits, but differ from the
diploid sexuals, then these traits can be said to be directly affected by ploidy level. If,
however, the hybrid genotypes are in-between the sexuals and the established
61
apomicts, it seems likely that the sexual mother and the apomictic father have
contributed different alleles to the offspring. It can then be argued that selection has
favoured different alleles, hence different phenotypes, in sexuals and apomicts.
Systems containing both sexuals and apomicts can be stable, despite the more
efficient reproduction of apomicts, if niche differentiation between the to forms exists.
If polyploids have different characteristics from diploids because of their polyploidy,
this can stabilise the system without any further effects of reproductive mode.
Therefore, our first question is whether polyploidy causes differences that can lead to
niche differentiation between sexuals and apomicts. Our second question is whether
apomicts are under selection when they establish, and on which characters selection
acts. Selection can be the consequence of competition with conspecifics and may lead
to niche shifts. Other forces exerting selection may be specific for asexual
reproduction, e.g. favouring general-purpose genotypes (Lynch 1984;De Kovel & De
Jong 2000), or in some other way be connected to asexuality.
We compared morphological and life-history traits in sexuals, apomicts and
their triploid hybrids to study the effect of polyploidy and selection on those traits. In a
previous study comparing hybrids and established apomicts it was shown that
apomicts are probably selected for a diverging phenology and longer leaves, in
particular under shaded conditions. In the present study, growth and development
were studied more closely, as well as reproduction-related traits. Also, in the present
study, diploid sexuals were included, ascertaining differences between diploid sexuals
and triploid apomicts.
Material and methods
Seeds from 'new' and 'established' apomicts were collected from a generation
of plants grown in the greenhouse. For the origin of the new apomicts, see De Kovel
& De Jong, 2000. In spring 1998, seeds of sexual plants were collected from the field
that had also provided the 'mothers' for the new triploids. Seeds germinated in petridishes and were planted in 12x12 cm pots, filled with a 3:1 black soil:sand mixture
that were placed in the greenhouse on 19 July 1999. In total, seeds from 5 diploid, 7
established triploid, and 8 new triploid mothers were used. Each mother was
represented by three plants. In case of the apomicts, these were likely to be the same
genotypes; in the case of the sexual plants, these were probably half-sibs. Every week,
number of leaves and the length of the longest leaf were scored. In addition, height of
the highest reaching leaf was recorded on 13 September and 4 October. Size of
stomatal cells was measured on ten cells of a mature leaf. On 25 October, all plants
were transferred to an open greenhouse, so as to experience normal seasonal changes.
When plants started to flower, the date of flowering of each capitulum was
recorded. As few insects are present in the greenhouse, sexual plants were hand-
62
Chapter 5
pollinated with pollen from other sexual plants in the experiment. Seeds were
harvested and divided into developed and undeveloped seeds. Ovules per capitulum
were counted, and three sets of ten developed seeds were weighed. Thirty developed
seeds per capitulum were placed on wet filter paper in a petri dish and placed in an
incubator. After 14 days at 20°C in light, the number of seeds that had germinated was
counted.
Plants were harvested after flowering on 10 July. Leaf area per plant was
measured on the fresh leaves. Leaves and roots were dried for 48h at 70°C and
weighed. Leaf area of the whole plant was determined. Specific leaf area (SLA) was
calculated as total leaf area per plant / total dry weight of leaves.
Data analysis
Differences between classes in cell size and in leaf height were analysed with
an ANOVA with class as a fixed factor and mother as a random factor nested within
class. The same test was used for dry weights and specific leaf area (SLA) after a logtransformation to improve normal distribution of these data. SLA was calculated as
the total leaf area of the plant divided by total dry weight of the leaves.
The same kind of ANOVA but with individual plant as a random factor nested
within mother was used for data with a number of observations per plant. These were
the number of ovules per capitulum, and weight of seeds, as well as the proportions of
developed and germinated seeds after arcsine transformation. For post-hoc tests the
method of Student-Newman-Keuls (SNK) was used.
To analyse leaf length and leaf number a similar ANOVA was used with date
of observation as a random factor.
Pearson correlations were calculated for the relation between seed weight and
germination probability, seed weight and capitulum sequence number, seed weight
and the number of ovules per capitulum, number of capitula and number of ovules per
capitulum, SLA and cell size, and for total dry weight and number of capitula per
plant. The correlation between leaf length and leaf number was corrected for date.
Differences between classes in the number of capitula per plant and in the
appearance date of capitula were analysed with Kruskall-Wallace (K-W) nonparametric test.
Heritability Estimates
The heritability of some traits was estimated from the added variance
component for genotypes in an ANOVA design, following the method of Falconer
(Falconer 1981). The variance components were estimated from the mean squares
63
estimates in a Type I nested ANOVA design (SPSS software) with mothers as a
random factor. Significant variance components (p<0.05) for mothers were interpreted
as heritability values differing from zero.
Heritability of leaf length and leaf number for each class, each date was
inferred from significant effects of mother in an ANOVA with mother as a random
factor.
Results
General
One plant died during the experiment; it was a hybrid.
Fig. 1. Average leaf numbers through time of
sexual, apomictic and hybrid plants. Error
bars show one standard error. Open circles:
established apomicts; closed circles: hybrids;
squares: sexuals.
Fig. 2. Average length of the longest leaf
through time of sexual, apomictic and
hybrid plants. Error bars show one
standard error. Open circles: established
apomicts; closed circles: hybrids; squares:
sexuals.
Growth and Morphology
Cell sizes differed significantly between the different classes (p<0.001). The
stomatal cells of the sexuals were shorter than those of the apomicts and hybrids by
10.6% on average, but apomicts and hybrids did not differ significantly.
64
Chapter 5
Fig. 3. The average number of ovules per
capitulum vs. the number of capitula per plant
for sexual, apomictic and hybrid plants. The
line shows the linear regression fit.
The number of leaves per plant was significantly different between the
different classes (p<0.001). Sexuals had most leaves whereas hybrids had on average
fewest leaves, though this difference between hybrids and apomicts was mainly
apparent in autumn and spring, when leaf numbers were high (Fig. 1). Leaf length
differences varied with the time of year (class * time interaction p<0.001): sexuals had
shorter leaves in the autumn when leaves were long, but longer leaves in spring when
leaves were short. In autumn, hybrid leaf length was in-between sexuals and apomicts;
in spring it was close to the low value of the apomicts (Fig. 2). Not one class had
longer leaves than the others did on average (p=0.644). Timing of leaf growth, the
phenology, was not conspicuously different between the different classes.
Leaf height of apomicts was significantly higher than leaf height of sexuals
and hybrids on 13 September (p<0.001) by about 2 cm, but no significant differences
were found in leaf height on 4 October (p=0.395). Significant differences between
mothers were found for leaf height on both dates (p<0.01).
65
Fig. 4 The fraction of developed seeds
per capitulum in sexuals, apomicts and
hybrids. The boxes contain 50% of the
data, the fat line shows the median, and
whiskers extend from lowest to highest
value excluding outliers.
Fig. 5 The probability of germinating vs.
the average seed weight per capitulum.
Line shows the linear regression fit. Open
circles: established apomicts; closed
circles: hybrids; squares: sexuals.
Flowering and Seed Production
Of all surviving plants, only one did not flower; this was a hybrid.
The flowering plants produced 4.4 (±1.6) capitula per plant, and the numbers
did not differ significantly between different classes of plants (p=0.599). The date of
appearance of the first capitulum per plant was not significantly different between the
classes (p=0.075), but timing of all capitula differed significantly: hybrid flowers
appeared earliest and sexual flowers last (p<0.001). This was the same as the trend
found for first capitula. Sexuals produced on average 174±37 ovules per capitulum,
hybrids 158±32, and apomicts 147±33 (p=0.425). We estimated the total number of
ovules as the number of capitula times the average number of ovules per capitulum per
plant. This total number of ovules per plant was significantly higher in sexuals than in
apomicts and hybrids, 980 (±576) in sexuals, and 608 (±149) and 693 (± 223) in
apomicts and hybrids respectively (p=0.034). This difference in ovule number per
plant, despite insignificant differences in capitulum number and ovule number per
capitulum, was the result of a striking pattern of variation. In sexuals we found a
positive correlation between the number of capitula per plant and the number of
ovules per capitulum (r2=0.42, p=0.023), whereas in apomicts we found a negative
correlation (r2=0.20, p=0.026). In the hybrids, no significant correlation was found
(r2=0.01, p=0.730). So, in sexuals, plants with many capitula also had many ovules per
66
Chapter 5
capitulum, whereas in apomicts plants with many capitula had few ovules per
capitulum (Fig. 3).
Though the established apomicts invaria bly had a high proportion of
developed seeds (0.94±0.06), the proportion in the hybrids varied much more and the
average was lower than in the apomicts (0.71±0.22) (p<0.001) (Fig. 4). Sexuals, too,
had often a low proportion of developed se eds (0.41±0.23), and in a number of cases
the seedhead had not developed at all. However, this was probably due to pollen
limitation in the greenhouse, since hand-pollination is not completely efficient.
The fraction of the mature-looking seeds that germinated was 68% (± 31) and
69% (±28) in the sexuals and established apomicts respectively. In the hybrid
apomicts, the germination was 54% (±30). These differences were not significant
(p=0.223). Variation between petri dishes, though, was rather high, because some
dishes became infected by fungus.
Seed weight of mature-looking seeds decreased with capitulum sequence
number (p<0.001). Though the number of capitula was the same in all three types, the
variation in capitulum number varied among the types and this complicated the
analysis of seed weight. Over all capitula, without taking capitulum number into
account, seed weight did not differ significantly among the different classes (p=0.527).
Heavier seeds had a higher probability of germinating (p<0.001) (Fig. 5). This
relationship did not differ between the classes (p=0.537).
Seed weight decreased significantly with increasing number of ovules per
capitulum, but r2 was low (r2 =0.04, p=0.011 n=159).
Harvest
Total dry weight was 4.68 g (±1.57) per plant on average and did not differ
significantly between classes (p=0.169), though weight of taproots separately was
significantly higher in the sexuals than in the two triploid classes (p=0.016). Leaf area
was on average 110 cm2 (± 60) and not significantly different between classes
(p=0.887). Specific leaf area (272 ± 43 cm2 g-1) was the same for all classes as well.
The total dry weight did not correlate with the number of capitula in sexuals or
apomicts (p=0.842 and 0.270 resp.), but did so in the hybrids (p<0.001, r2=0.58) (Fig.
6).
Specific leaf area was not dependent on the average cell size (p=0.678).
Trait Heritability
67
Fig. 6. The number of capitula
per plant vs. the total dry
weight (g) at harvest for
sexuals, apomicts and
hybrids. Lines show linear
regression fit.
For selection to work, traits must be heritable. With the set-up that we used,
an estimate of heritability of some of the traits could be made. For apomicts,
heritability estimates are broad-sense, for sexuals narrow-sense. Hybrids had
heritability values larger than zero for the number of seeds per capitulum, the fraction
of seeds that developed, the fraction of seeds that germinated and the weight of seeds
in the first capitulum. For the apomicts only the number of seeds per capitulum had a
heritability significantly different from zero. For all these traits, heritability values of
the hybrids were higher than those of the apomicts (Table 1).
Leaf length and leaf number had been measured 37 times on the same plants.
A significant effect (p<0.05) of mother on leaf length was found 3 times in the
sexuals, 10 times in the apomicts and 13 times in the hybrids. These positive values
were found mainly from January until May, when leaves are relatively short. A
significant effect of mother on leaf number was found 12 times in the sexuals and 19
times in the apomicts, mainly from November until March when leaf numbers were
relatively small. For the hybrids we found a significant effect of mother on all dates
but four.
68
Chapter 5
2
Table 1. Heritability (h ) estimates of seed-related traits. Heritability of traits in sexuals is
calculated as though offspring were full sibs, though the seedheads probably contained a
mixture of full and half-sibs. An asterix denotes significant differences from zero.
Trait
ovules per capitulum
fraction ovules developed
germination fraction
seed weight first
capitulum
Sexuals
0.14
Apomicts
0.22*
Hybrids
0.44*
-
0.20
0.67*
0.57
0.18
0.29*
-
0.06
0.51*
Discussion
The comparison between sexual and apomictic Taraxacum and their hybrids
resulted in different patterns for different traits. The different patterns and their
relevance will be discussed below.
No Differences
Traits that did not differ significantly between all classes of Taraxacum were
capitula number and seed weight, when viewed over all capitula. Capitula number was
not shown to differ between sexuals and apomicts in previous experiments either (De
Kovel & De Jong 1999). In a field survey in Central Europe, however, sexual
Taraxacum were found to have smaller and more numerous capitula than apomicts in
the same field (Den Nijs et al. 1990). In the current study, sexuals on average
produced more ovules per plant than the apomicts. This was connected to a particular
pattern of variation in capitula number and ovule number (Fig. 3), and we can
therefore not easily interpret this as an adaptation of apomicts to their reproductive
assurance, as Den Nijs et al. do for their data.
Seed weight did not differ between the classes in this experiment, nor in a
previous experiment that compared sexuals and apomicts (De Kovel & De Jong 1999).
Hybrids Traits Identical to Apomicts, Ploidy Effect
Cell size did not differ between hybrid and established apomicts, but was
smaller in sexuals. It seems clear that triploidy causes larger cell sizes than diploidy.
This is commonly found (Tal 1980;Levin 1983) and may affect further physiology of
the plants (Warner & Edwards 1993).
69
Sexuals had a ± 40% heavier taproot than both apomicts and hybrids, a pattern
not found in a previous comparison between sexuals and apomicts (De Kovel & De
Jong 1999).
For much of the season sexuals had considerably more leaves than the two
triploid classes. In spring and autumn, hybrids had even fewer leaves than the
established apomicts. It is likely that triploidy caused lower leaf numbers.
Hybrid Traits not in-between Sexuals and Apomicts, Selection on Nonadditive Traits
One complication is that newly formed apomictic triploids can have
development errors. Such errors have been found in other sexual-asexual hybrids
(Wetherington et al. 1987), as well as in sexual-sexual inter-generic hybrids. It is well
possible that there are strong epistatic effects rather than additive effects for some
traits. In that case only some combinations of sexual and apomictic genomes produce
fit phenotypes, and hybrids are not in-between their parents.
One trait in which problems obviously occur is in seed production. Seed
production of newly formed apomicts was poor in many cases. Though capitula were
formed and ovules were formed in those capitula, parthenogenetic development of
seeds was problematic. It has been shown in hybrid studies that fertility is often more
vulnerable than vigour (Forsdyke 2000) (Coyne & Orr 1989). It is possible that
fertility-related traits show strong epistasis (Merila & Sheldon 1999).
Even more complicated is the fact that hybrids had on average fewer leaves
than either sexuals or established apomicts, though this was only apparent in some
seasons. Should this be attributed to general developmental problems, or is this a
polyploidy effect on which selection subsequently acts towards more leaves?
Similarly, hybrids flowered earlier than sexuals or established apomicts. It has
been shown before that the cue to which plants react for timing of flowering is
probably complex and sensitive, because changes in conditions can reverse the order
of flowering of different groups of plants (Segraves & Thompson 1999;De Kovel &
De Jong 1999). Therefore, it is difficult to conclude about the direction of ploidy
effect or selection for the field situation, though both may play a role.
Hybrid Traits in-between Sexuals and Apomicts, Selection on Additive
Traits
Hybrid leaf length was in-between sexuals and established apomicts in winter,
when leaves are longest and light levels are low. In an earlier experiment, it was
shown that the hybrids had shorter leaves than established apomicts when grown in
70
Chapter 5
the shade, though not in full light (De Kovel & De Jong 2000). Selection for leaf
elongation in shade is a likely explanation for this pattern. A comparison of sexuals
and apomicts collected from a single field also showed that apomicts had a stronger
leaf length response to shading than sexuals (De Kovel & De Jong 1999). In a field
survey in Central Europe, sexual Taraxacum were found to have smaller and narrower
leaves than apomicts in the same field (Den Nijs et al. 1990). This suggests that the
pattern is widespread.
A funny ‘trait’, the correlation between number of seeds per capitulum and
number of capitula per plant, also, was in-between sexuals and established apomicts
for the hybrids. This suggests a genetic component in this correlation that is different
in sexuals and apomicts, and possibly under selection. However, the interpretation of
this pattern is difficult.
Heritabilities
Selection is only effective in changing trait values, if the trait values are
heritable. Since we want to see whether there is selection on hybrids, we are, of
course, especially interested in the heritability of specific traits in the hybrids.
Selection may have reduced the heritabilities in the established apomicts, and even in
the sexuals.
The fraction of developed and germinable seeds had a positive heritability in
the hybrids, but not so in established apomicts, and positive selection on these traits
probably took place. Seed weight had a positive h2 in hybrids and a lower h2 value in
the established apomicts. Selection could take place, and probably did, but selection
did not clearly act towards heavier seeds, despite the correlation between seed weight
and germination ability. Ovule number per capitulum was also a heritable trait with a
higher h2 value in apomicts than in hybrids. Possibly, the weak correlation between
ovule number and seed weight caused balancing selection on both traits (Tweney &
Mogie 1999).
Leaf length was heritable in the hybrids as well, but in particular in spring,
when leaf lengths were the same in established apomicts and hybrids. Leaf number
had positive h2-values in the hybrids in most of the season. Established apomicts had
positive h2 values except in autumn and spring, which tentatively is evidence for
selection towards higher leaf numbers in those periods. Heritability values at
subsequent dates are, of course, not independent. I have calculated them as if they
were, to get a rough indication of the process that is going on.
71
Niche Differentiation because of Ploidy Differences
Triploidy caused larger cell sizes. Cell size and shape can affect
photosynthetic rate per unit leaf area in either direction (Warner & Edwards 1993).
The larger taproots in sexuals may be a result of higher photosynthetic rates. Whether
the cell size difference causes niche differentiation between sexual and apomictic
Taraxacum is not clear from this experiment, though it is conceivable. Triploidy
probably caused apomicts to have fewer leaves than sexuals, though this pattern was
less clear-cut. The ecological significance of fewer leaves is not quite clear. It is
possible that having many smaller leaves results in more economic water use (Dudley
1996). This in turn could result in niche differences with sexuals having advantage in
localities that are more arid. Surveys of distribution of sexual and apomictic
Taraxacum also show a more arid preference of sexuals (Roetman et al. 1988). It
seems likely that polyploidy also influences the onset of flowering, but sexual and
apomictic populations still show a large overlap in flowering time. It seems therefore
unlikely that this causes enough niche differentiation to allow co-existence of the two
types.
Selection on New Apomictic Lineages
Apomictic lineages were selected for higher proportions of developed,
germinable seeds. These traits are directly connected to fitness, so this is not
surprising. Such selection also does not cause niche differences between the sexuals
and apomicts. Despite heritable variation in the hybrids for seed weight and number of
ovules per capitulum, selection did not noticeably act towards trait values differing
from those in sexuals. Apomicts were probably selected for a more plastic leaf length
response to shading. This may enable the apomicts to grow in locations with more
shade or with more variable light conditions than the sexuals can. Selection for higher
leaf plasticity may be explained as an adaptation to lower genetic variation in the
offspring. It is also possible that the effects of triploidy on leaf number and cell size
need to be balanced by changes in other leaf characteristics to ensure optimal growth.
As a side effect, this may increase niche differences between sexuals and apomicts.
From this study, there is no clear evidence for selection for traits specifically
connected with the apomictic mode of reproduction. Differences between sexuals and
apomicts in Taraxacum in the traits under study are likely to be mainly the
consequence of ploidy differences. Repeated formation of new clones, and a short
lifetime of clonal lineages may be the reason that differences between sexuals and
apomicts are relatively small.
72
Chapter 5
Acknowledgements
I thank Peter van Dijk of the Nederlandse Instituut voor Oecologisch
Onderzoek (NIOO-CTO) for giving me the seeds of the established apomicts and the
hybrids. I am grateful to Z. Bochdanovits and G. de Jong for useful comments on
earlier drafts of this manuscript. This work was supported by the Life Sciences
Foundation (SLW), which is subsidised by the Netherlands Organisation for Scientific
Research (NWO).
73